There are a few meanings of a "successful" action, but I can answer in several ways.
If the goal is to find measures that a computer can easily evaluate as good or bad, then one example would be paying debt with cash while someone else holds your stock. Since buying back stock before paying debt is preferable, paying the debt first indicates that a player is in desperate straits. If a player has cash reserves, then losing in the market is also an indicator of poor play: if someone buys food and you buy food right after, then you should have bought food first. Losing on sells is an indicator of poor play regardless of cash reserves. But care should be taken that such losses are based on true market pressure; if a player buys and sells 100 food every half second, then this would indicate "poor play" for every other player holding food, having lost on every single trade. Clearly, no such poor play exists, and this price manipulation is all artificial, with no actual effects.
If the goal is to evaluate a player's strength, then a good indication is moving into an industry before everyone else. Players usually only think about something like power or water when they see "uh oh, power is at $300, time to build solar panels". But recognizing the upward trend beforehand is more uncommon, and better players can grasp the future better. It's an even stronger play when the resource can be stockpiled, such as food: get into food first, build a stockpile if you don't have pressing cash needs, and then sell the stockpile after everyone else moves into farms. But this skill becomes less important as ROI gets higher, because excess cash goes into investment instead of winning markets. Another indication is how much time they spend looking at everyone else's stuff, instead of their own stuff.
If the goal is to improve one's own strength, then the most I can say is looking for arbitrage. This could be simple, such as if an Offworld makes $500/s vs an additional PD at $600/s. Or, if buying a player yields 100%/day ROI, and a transition yields only 80%. But it could be more complex, with opportunity costs of eliminating players, or exclusive patents, etc. The arbitrage then revolves around significant timepoints, such as stock-buying time/tile-grabbing time/patent time, and then the worth of everything is decided by these timepoints, adjusted appropriately. But this is more of an AI issue; humans find it difficult to do such calculations.
